Simulating Evolution in Competitive AI Agents Shows that Novel Survival Stresses Like Pandemics can Lead to Evolutionary Weakening

It may be expected that survival stresses lead to evolutionary strengthening of the population. By eliminating the weakest individuals, stress can lead to increase in strength of the average individual. However, this may not always be true. Through evolutionary simulation of AI agents, we show that novel survival stress factors, especially those focused on narrow characteristics, can even reduce the strength of the population in other characteristics. When a novel survival stress appears on a narrow set of characteristics, it eliminates the individuals weak in those characteristics, individuals who would have been strong in other characteristics to compensate for the weakness in stressed attributes. Thus, the availability of strength in the non-stressed characteristics in the population can get reduced. To quantify and evaluate the significance of this effect, we use AI agents to simulate evolution under novel survival stress factors. Evolutionary machine learning algorithms have been used to simulate biological evolution. The implication of this effect for evolutionary machine learning is also discussed.